Security audit

Memory V2

Security checks across malware telemetry and agentic risk

Overview

This is a coherent memory system, but it can persist, transform, and optionally send private long-term memories to an LLM without enough privacy and retention guidance.

Review before installing if memories may contain personal data, secrets, customer records, or private conversations. Use an isolated database path, avoid storing raw transcripts by default, add redaction and retention rules, control or disable LLM summarization unless the provider is trusted for that data, and keep backups before enabling automatic maintenance.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (5)

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The documentation explicitly demonstrates storing raw user conversation content and extracting named personal entities, but provides no notice about consent, retention, minimization, or privacy obligations. In a memory system, this omission can lead integrators to persist sensitive personal data by default, increasing privacy, compliance, and misuse risk.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The migration example instructs users to copy all historical memories into the new system without warning about sensitive legacy data, duplication, or persistence scope. This can cause unreviewed personal or confidential information to be re-ingested and retained in a new store, potentially broadening exposure and violating retention expectations.

Missing User Warnings

Medium

Confidence: 83% confidence
Finding: The delete and maintenance sections describe deletion, compression, archiving, and forgetting behavior without warning users about data lifecycle consequences. This can mislead integrators into assuming deletion semantics are straightforward or reversible, and may result in unexpected data loss, residual copies, or noncompliant retention handling.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The skill explicitly supports persistent memory storage, deletion, automatic forgetting, and compression, but the documentation does not clearly warn users about privacy exposure, retention behavior, or irreversible data loss. In a memory system, these omissions matter because users may store sensitive personal information and assume stronger safeguards or recoverability than actually exist.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The module sends raw memory content to a configured LLM for summarization, which can disclose potentially sensitive stored data to an external service. In a memory-management component, this is more dangerous because the content may include user history, secrets, or other long-lived context, and there is no consent gate, redaction step, or locality guarantee before transmission.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.